Unix vmstat 命令

VirtualMemoryStatistics(vmstat)是报告关于进程,虚拟内存,磁盘,CPU活动的一个统计报告.
在一个多CPU系统中,vmstat统计所有个数的CPU的平均输出.为每个进程进行统计.没有参数时,vmstat显示每一行是自从系统启动以后的虚拟内存的活动情况.

vmstat的语法我们可以通过linux的man命令来查看。

将man的结果导入到指定的文本

#manvmstat|col-b>/u01/tianlesoftware.txt

查看vmstat用法

#manvmstat

VMSTAT(8) LinuxAdministrator’sManual VMSTAT(8)

NAME

vmstat-Reportvirtualmemorystatistics

SYNOPSIS

vmstat[-a][-n][delay[count]]

vmstat[-f][-s][-m]

vmstat[-Sunit]

vmstat[-d]

vmstat[-pdiskpartition]

vmstat[-V]

DESCRIPTION

vmstatreportsinformationaboutprocesses,memory,paging,blockIO,

traps,andcpuactivity.

Thefirstreportproducedgivesaveragessincethelastreboot. Addi-

tionalreportsgiveinformationonasamplingperiodoflengthdelay.

Theprocessandmemoryreportsareinstantaneousineithercase.

Options

The-aswitchdisplaysactive/inactivememory,givena2.5.41kernel

orbetter.

The-fswitchdisplaysthenumberofforkssinceboot. Thisincludes

thefork,vfork,andclonesystemcalls, andisequivalenttothe

totalnumberoftaskscreated.Eachprocessisrepresentedbyoneor

moretasks,dependingonthreadusage.Thisdisplaydoesnotrepeat.

The-mdisplaysslabinfo.

The-nswitchcausestheheadertobedisplayedonlyonceratherthan

periodically.

The-sswitchdisplaysatableofvariousevent countersandmemory

statistics.Thisdisplaydoesnotrepeat.

delayisthedelaybetweenupdatesinseconds. Ifnodelayisspeci-

fied,onlyonereportisprintedwiththeaveragevaluessinceboot.

countisthenumberofupdates. Ifnocountisspecifiedanddelayis

defined,countdefaultstoinfinity.

The-dreportsdiskstatistics(2.5.70oraboverequired)

The-pfollowedbysomepartitionnamefordetailedstatistics(2.5.70

oraboverequired)

The-SfollowedbykorKormorMswitchesoutputsbetween 1000,

1024,1000000,or1048576bytes

The-Vswitchresultsindisplayingversioninformation.

FIELDDESCRIPTIONFORVMMODE

Procs

r:Thenumberofprocesseswaitingforruntime.

b:Thenumberofprocessesinuninterruptiblesleep.

Memory

swpd:theamountofvirtualmemoryused.

free:theamountofidlememory.

buff:theamountofmemoryusedasbuffers.

cache:theamountofmemoryusedascache.

inact:theamountofinactivememory.(-aoption)

active:theamountofactivememory.(-aoption)

Swap

si:Amountofmemoryswappedinfromdisk(/s).

so:Amountofmemoryswappedtodisk(/s).

IO

bi:Blocksreceivedfromablockdevice(blocks/s).

bo:Blockssenttoablockdevice(blocks/s).

System

in:Thenumberofinterruptspersecond,includingtheclock.

cs:Thenumberofcontextswitchespersecond.

CPU

ThesearepercentagesoftotalCPUtime.

us:Timespentrunningnon-kernelcode.(usertime,includingnicetime)

sy:Timespentrunningkernelcode.(systemtime)

id:Timespentidle.PriortoLinux2.5.41,thisincludesIO-waittime.

wa:TimespentwaitingforIO.PriortoLinux2.5.41,shownaszero.

FIELDDESCRIPTIONFORDISKMODE

Reads

total:Totalreadscompletedsuccessfully

merged:groupedreads(resultinginoneI/O)

sectors:Sectorsreadsuccessfully

ms:millisecondsspentreading

Writes

total:Totalwritescompletedsuccessfully

merged:groupedwrites(resultinginoneI/O)

sectors:Sectorswrittensuccessfully

ms:millisecondsspentwriting

IO

cur:I/Oinprogress

s:secondsspentforI/O

FIELDDESCRIPTIONFORDISKPARTITIONMODE

reads:Totalnumberofreadsissuedtothispartition

readsectors:Totalreadsectorsforpartition

writes:Totalnumberofwritesissuedtothispartition

requestedwrites:Totalnumberofwriterequestsmadeforpartition

FIELDDESCRIPTIONFORSLABMODE

cache:Cachename

num:Numberofcurrentlyactiveobjects

total:Totalnumberofavailableobjects

size:Sizeofeachobject

pages:Numberofpageswithatleastoneactiveobject

totpages:Totalnumberofallocatedpages

pslab:Numberofpagesperslab

NOTES

vmstatdoesnotrequirespecialpermissions.

Thesereportsareintendedtohelpidentifysystembottlenecks. Linux

vmstatdoesnotcountitselfasarunningprocess.

Alllinuxblocksarecurrently1024bytes.Old kernelsmayreport

blocksas512bytes,2048bytes,or4096bytes.

Sinceprocps3.1.9,vmstatletsyouchooseunits(k,K,m,M)default

isK(1024bytes)inthedefaultmode

vmstatusesslabinfo1.1 FIXME

FILES

/proc/meminfo

/proc/stat

/proc/*/stat

SEEALSO

iostat(1),sar(1),mpstat(1),ps(1),top(1),free(1)

BUGS

Doesnottabulatetheblockioperdeviceorcountthenumberof sys-

temcalls.

AUTHORS

WrittenbyHenryWare<[e-mail]>.

FabianFrédérick<[e-mail]>(diskstat,slab,partitions...)

ThroatwobblerGinkgoLabs 27July1994 VMSTAT(8)

示例：

#vmstat210

procs-----------memory---------------------------swap--------io-------system-------cpu------

rbswpdfreebuffcachesisobiboincsussyidwa

11208289603812826820400278401050370546445

1020828896381282682040012801054310438581

20208288963813626819600257551056301439525

202082489638136268196002571104259368770

00208286723813626819600257251052372446455

002082882438140268192001291111052316332641

102082876838148268184002571091050286970211

1020828512381482681840025711046327648432

0020828832381482681840025711047313542521

002082883238152268180002571091050308433595

每2秒显示一次，共显示10次。

在Unix(Solaris)平台下：

-bash-3.00$vmstat210

kthrmemorypagediskfaultscpu

rbwswapfreeremfpipofrdesrs0s1----insycsussyid

00050163444200040126512350000010073114167233097

0002065576237511214540000000008706259503096

00020655122375048115000000000095152910663097

0002065512237504811530000000008525489343097

000206551223750484400000000008364839073097

0002065512237504814460000000008736329583097

00020655122375048000000000008124178633097

000206551223750484400000000008104818743097

00020653842374920000000001008014068623096

0002065384237492014460000000008175458983097

<!--EndFragment-->

一些标准说明：

CPU:

1)如果(processr)是连续的大于在系统中的CPU的个数表示系统现在运行比较慢,有多数的进程等待CPU.
2)如果r的输出数大于系统中可用CPU个数的4倍的话,则系统面临着CPU短缺的问题,或者是CPU的速率过低,系统中有多数的进程在等待CPU,造成系统中进程运行过慢.
3)如果空闲时间(cpuid)持续为0并且系统时间(cpusy)是用户时间的两倍(cpuus)系统则面临着CPU资源的短缺.

简单的说就是：

r和b不高于5，如果r经常大于4，且id经常少于40，则表示CPU负荷很重。

us+sy不超过80%，如果大于80，进程就会在运行队列中花费等待时间，响应时间和吞吐量就会下降。

wa不超过25，如果说wa>40表明磁盘io没有也许存在不合理的平衡，或者对磁盘操作比较频繁

一般情况下，导致CPU资源不足都是应用程序的问题.比如一些SQL语句不合理等等都会造成这样的现象.

内存：

si和so基本为零。其中si表示当前每秒交换回内存（Swapin）的总量，单位为kbytes；so表示当前每秒交换出内存（Swapout）的总量，单位为kbytes。如果si，so长期不等于0，表示系统使用swap频繁，即内存不足。

一般情况下，如果内存的占用率比较高,但是,CPU的占用很低的时候,可以考虑是有很多的应用程序占用了内存没有释放.

要注意的是，Linux内存在使用后不会释放，所以一般会看到内存使用率很高，要判断linux内存是否正常，需要通过对swap交换区的交换频率来看。

具体参考blog：Linux内存机制

http://blog.csdn.net/tianlesoftware/archive/2010/04/08/5463790.aspx

<!--EndFragment-->